What happens when AI tools let you adjust the style of a result instead of just typing a better prompt?
How do style lenses and output shaping change how users interact with AI results?
This looks at what happens when AI interfaces give users controls for steering the form of a result (style lenses, previews, sliders, canvases) instead of only a chat box, and how the look and polish of AI output changes the way people judge and act on it.
This looks at two sides of one thing: the controls that let users steer the form of an AI result, and the way that form shapes what users do with the result. The term "style lenses" comes from Friedman, who names output shaping as one of five design levers that move AI interfaces away from slow, typed chat. The others are intent capture, refinement, agentic tasks and workflow embedding. The pitch is simple. Rewriting a prompt to say "make it shorter, more formal" is clumsy. A lens or slider lets you reshape output directly, the way you'd adjust a photo filter Can task-oriented UI patterns reduce the friction of AI chat interfaces?. A related proposal goes one step further and asks for previews of the likely output *before* you submit a prompt. That way you catch a misunderstanding early instead of discovering it three conversational turns later Can previewing AI output before submission reduce prompt iteration?.
Behind this sits a less obvious point: the model never sees "your prompt" on its own. The IAI model describes what a generative AI actually receives as an "augmented instruction": your words plus whatever constraints the interface's sliders, selections and modes add. Across 66 tools, the authors find twelve recurring interaction patterns that build these instructions How do interaction signals reshape what generative AI actually receives?. So a style lens is not just decoration. It quietly rewrites the request. That matters because AI output is unstable by nature. It shifts with wording, context and sampling Why does AI output change with every prompt and context?. The context feeding it is also invisible and always changing, in a way that traditional software menus never were How does AI context differ from conventional software context?. Output-shaping controls help by making some of that hidden context visible and adjustable. Other parts stay locked away. System prompts set by developers form a hidden layer that users can't see or touch, and a small edit there can change behavior overnight Who actually controls what AI systems say?.
The other side is less comfortable: shaping how output looks also shapes how much people trust it. Polished, professional-looking output borrows an old shortcut, "this looks expert, so someone expert made it." Less experienced workers are the most exposed, because they can't judge what's underneath Does polished AI output trick audiences into trusting it?. Samsonov argues that chat interfaces make this worse. Confident tone and clean formatting create a halo effect, and the interface nudges you toward your next question instead of toward checking sources Does confident AI tone make users skip fact-checking?. Layout matters too. In an eye-tracking study, AI Overviews at the top of search results drew attention away from the first-ranked link, which fell from 31% to 9% of attention, and people trusted both equally Where do searchers look when AI Overviews appear?.
The twist you may not expect: polish can't be cleanly separated from distortion. Writers preferred AI rewrites 63% of the time, yet objected to the way those same rewrites shifted their personal voice. Mitigation studies suggest that, inside the model, the polish people like and the distortion they dislike come from the same optimization Can user preference guide AI writing tool alignment?. A style lens that makes output "better" may be turning up both at once. Separately, tools that generate interfaces skip roughly a quarter of the design reasoning they claim to follow Do generative UI tools actually implement their stated design rationales?. So a polished surface can promise more than it delivers.
The corpus is richer on the trust effects of polish than on studies of style-lens controls themselves. Friedman and the feedforward paper are design proposals, not user experiments. The open question running through all of it: do output-shaping controls give users real steering, or more convincing packaging?
Sources 11 notes
Friedman identifies five design levers—intent capture, output shaping, refinement, agentic tasks, and workflow embedding—that sidestep slow typed prompts. Examples include node-based intent visualization, style lenses for output, and canvas manipulation, shifting AI interfaces away from conversational chat toward direct interaction.
A preprint proposes that GenAI systems need feedforward design—showing anticipated output before prompt submission—to bridge the gulf between user intent and model response. Four key properties (anticipation, disambiguation, direct engagement, reflection) and a detailed LLM UI example demonstrate how live previews can surface misalignments and reduce excessive conversational exchanges.
The IAI model formalizes human-GenAI interfaces as a six-entity graph where Augmented Instruction—machine-readable encoding of prompts plus interaction constraints—is what GenAI actually receives. Applied to 66 tools, it distills twelve recurring, composable interaction paradigms and enables design comparison where prior taxonomies could not.
AI outputs exhibit essential mutability—they vary with sampling, prompt wording, and audience interpretation. This is not a defect but a defining feature of tokens as media, making them fundamentally different from fixed commodities and resistant to traditional quality assurance.
AI interactions operate on a substrate of constantly shifting context—prompt, history, retrieved data, hidden state—that users cannot internalize like traditional UIs. This structural mutability demands a new design discipline centered on context engineering rather than interface design.
Show all 11 sources
System prompts operate as a confidential three-tier control structure where developers set foundational rules users cannot see or influence. Evidence from 2025 incidents shows small prompt edits can abruptly shift model behavior, yet users remain excluded from this layer.
Generative AI produces visually sophisticated outputs without underlying judgment, leveraging the historical heuristic that professional-looking work signals expert thinking. This substitution is especially risky for less experienced workers who lack domain knowledge to evaluate substance beyond form.
Samsonov argues that AI's confident tone, correct grammar, and polished formatting create a halo effect that makes users trust accuracy without verification. The chat interface compounds this by offering no accuracy warnings and actively prompting users toward new questions instead of source-checking.
Eye-tracking data shows AI Overviews receive significantly longer fixation times, reducing attention to the first-ranked result from 31% to 9%. Trust ratings between AI Overviews and ranked results remained equally high despite this attention shift.
Writers prefer AI rewrites 63% of the time but object to systematic persona distortions those same rewrites introduce. Mitigation studies show polish and distortion are entangled at the model level—preference optimization produces both simultaneously.
A benchmark of 24 tasks across five tools found roughly 25% of design rationales go unimplemented, rising to 34% for functional requirements. Tools recognized only half the UX principles embedded in prompts.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- Interaction-Augmented Instruction: Modeling the Synergy of Prompts and Interactions in Human-GenAI Collaboration
- Feedforward in Generative AI: Opportunities for a Design Space
- Design Theater: Evaluating the Gap Between User-Facing Design Reasoning and Implementation in Generative UI Tools
- Prototyping Multimodal GenAI Real-Time Agents with Counterfactual Replays and Hybrid Wizard-of-Oz
- Prototyping with Prompts: Emerging Approaches and Challenges in Generative AI Design for Collaborative Software Teams
- Who Controls the Conversation? User Perspectives On Generative AI (LLM) System Prompts
- An HCI-Centric Survey and Taxonomy of Human-Generative-AI Interactions
- Probing How Users Interact with Turn-Level Design Frictions for AI Chatbots